Hugging Face's logo

hugodk-sch
/
aftonposten-6b-align-scan

PEFT
Safetensors
gptj
alignment-handbook
trl
dpo
Generated from Trainer
4-bit precision
bitsandbytes
Model card Files Community
hugodk-sch commited on
Commit
7aff5a4
•
1 Parent(s): 3079f10

Training in progress, step 500

Browse files
Files changed (8) hide show
  1. README.md +15 -27
  2. adapter_config.json +4 -4
  3. adapter_model.safetensors +1 -1
  4. all_results.json +16 -16
  5. eval_results.json +12 -12
  6. train_results.json +5 -5
  7. trainer_state.json +0 -0
  8. training_args.bin +1 -1
README.md CHANGED
@@ -20,15 +20,15 @@ should probably proofread and complete it, then remove this comment. -->
20
 
21
  This model is a fine-tuned version of [data/ap-gpt-j-6b-sft-qlora-04-08](https://huggingface.co/data/ap-gpt-j-6b-sft-qlora-04-08) on the hugodk-sch/aftonposten_title_prefs dataset.
22
  It achieves the following results on the evaluation set:
23
- - Loss: 0.4946
24
- - Rewards/chosen: 0.0246
25
- - Rewards/rejected: -0.0031
26
- - Rewards/accuracies: 0.5922
27
- - Rewards/margins: 0.0278
28
- - Logps/rejected: -37.5481
29
- - Logps/chosen: -33.7882
30
- - Logits/rejected: -2.0897
31
- - Logits/chosen: -2.0944
32
 
33
  ## Model description
34
 
@@ -57,27 +57,15 @@ The following hyperparameters were used during training:
57
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
58
  - lr_scheduler_type: cosine
59
  - lr_scheduler_warmup_ratio: 0.1
60
- - num_epochs: 4
61
 
62
  ### Training results
63
 
64
- | Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
65
- |:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
66
- | 0.4975 | 0.26 | 100 | -2.2364 | -2.2316 | -33.8905 | -37.4088 | 0.4992 | 0.5307 | 0.0144 | 0.0036 | 0.0108 |
67
- | 0.4925 | 0.52 | 200 | -2.2354 | -2.2306 | -33.7318 | -37.2691 | 0.4987 | 0.5718 | 0.0303 | 0.0055 | 0.0248 |
68
- | 0.4825 | 0.78 | 300 | -2.2326 | -2.2278 | -33.7455 | -37.2678 | 0.4990 | 0.5453 | 0.0289 | 0.0040 | 0.0249 |
69
- | 0.4682 | 1.04 | 400 | 0.4981 | 0.0406 | 0.0330 | 0.5914 | 0.0076 | -37.1866 | -33.6285 | -2.2114 | -2.2163 |
70
- | 0.4532 | 1.3 | 500 | 0.4976 | 0.0426 | 0.0327 | 0.5777 | 0.0099 | -37.1895 | -33.6082 | -2.1796 | -2.1844 |
71
- | 0.4615 | 1.56 | 600 | 0.4960 | 0.0424 | 0.0253 | 0.5764 | 0.0172 | -37.2639 | -33.6102 | -2.1513 | -2.1561 |
72
- | 0.4447 | 1.82 | 700 | 0.4962 | 0.0323 | 0.0156 | 0.5943 | 0.0167 | -37.3605 | -33.7117 | -2.1360 | -2.1407 |
73
- | 0.4271 | 2.08 | 800 | 0.4951 | 0.0345 | 0.0124 | 0.5972 | 0.0221 | -37.3928 | -33.6895 | -2.1274 | -2.1322 |
74
- | 0.4126 | 2.34 | 900 | 0.4949 | 0.0293 | 0.0043 | 0.5943 | 0.0250 | -37.4733 | -33.7411 | -2.1074 | -2.1121 |
75
- | 0.4408 | 2.6 | 1000 | 0.4944 | 0.0272 | -0.0013 | 0.6092 | 0.0284 | -37.5294 | -33.7629 | -2.0966 | -2.1013 |
76
- | 0.4419 | 2.86 | 1100 | 0.4947 | 0.0250 | -0.0032 | 0.6213 | 0.0282 | -37.5488 | -33.7850 | -2.0910 | -2.0956 |
77
- | 0.4146 | 3.12 | 1200 | 0.4944 | 0.0258 | -0.0031 | 0.6038 | 0.0289 | -37.5475 | -33.7762 | -2.0897 | -2.0944 |
78
- | 0.4103 | 3.38 | 1300 | 0.4943 | 0.0262 | -0.0032 | 0.6213 | 0.0294 | -37.5482 | -33.7725 | -2.0900 | -2.0947 |
79
- | 0.4266 | 3.64 | 1400 | 0.4944 | 0.0253 | -0.0034 | 0.5951 | 0.0287 | -37.5508 | -33.7813 | -2.0897 | -2.0944 |
80
- | 0.416 | 3.9 | 1500 | 0.4943 | 0.0256 | -0.0040 | 0.6067 | 0.0297 | -37.5568 | -33.7782 | -2.0897 | -2.0944 |
81
 
82
 
83
  ### Framework versions
 
20
 
21
  This model is a fine-tuned version of [data/ap-gpt-j-6b-sft-qlora-04-08](https://huggingface.co/data/ap-gpt-j-6b-sft-qlora-04-08) on the hugodk-sch/aftonposten_title_prefs dataset.
22
  It achieves the following results on the evaluation set:
23
+ - Loss: 24.8075
24
+ - Rewards/chosen: -0.0054
25
+ - Rewards/rejected: -0.0092
26
+ - Rewards/accuracies: 0.5133
27
+ - Rewards/margins: 0.0038
28
+ - Logps/rejected: -37.6085
29
+ - Logps/chosen: -34.0885
30
+ - Logits/rejected: -2.2234
31
+ - Logits/chosen: -2.2282
32
 
33
  ## Model description
34
 
 
57
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
58
  - lr_scheduler_type: cosine
59
  - lr_scheduler_warmup_ratio: 0.1
60
+ - num_epochs: 1
61
 
62
  ### Training results
63
 
64
+ | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
65
+ |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
66
+ | 23.9156 | 0.26 | 100 | 24.9323 | -0.0015 | -0.0033 | 0.5282 | 0.0018 | -37.5501 | -34.0500 | -2.2289 | -2.2338 |
67
+ | 22.4879 | 0.52 | 200 | 24.7992 | -0.0025 | -0.0061 | 0.5598 | 0.0037 | -37.5780 | -34.0593 | -2.2258 | -2.2307 |
68
+ | 19.8329 | 0.78 | 300 | 24.9557 | -0.0066 | -0.0089 | 0.5216 | 0.0023 | -37.6060 | -34.1010 | -2.2229 | -2.2278 |
 
 
 
 
 
 
 
 
 
 
 
 
69
 
70
 
71
  ### Framework versions
adapter_config.json CHANGED
@@ -19,13 +19,13 @@
19
  "rank_pattern": {},
20
  "revision": null,
21
  "target_modules": [
22
- "o_proj",
23
- "up_proj",
24
  "q_proj",
 
 
25
  "v_proj",
26
  "k_proj",
27
- "down_proj",
28
- "gate_proj"
29
  ],
30
  "task_type": "CAUSAL_LM",
31
  "use_rslora": false
 
19
  "rank_pattern": {},
20
  "revision": null,
21
  "target_modules": [
22
+ "gate_proj",
 
23
  "q_proj",
24
+ "up_proj",
25
+ "o_proj",
26
  "v_proj",
27
  "k_proj",
28
+ "down_proj"
 
29
  ],
30
  "task_type": "CAUSAL_LM",
31
  "use_rslora": false
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:814fce2676951a9a2d42b13c5a1d1ab443de895adae7a07d14ec0f71f3860424
3
  size 176183216
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a3d6f37fbacb17abbce722857199ffab429b80c9da42b0ac74f2bd2e26a0bae0
3
  size 176183216
all_results.json CHANGED
@@ -1,21 +1,21 @@
1
  {
2
- "epoch": 4.0,
3
- "eval_logits/chosen": -2.0943572521209717,
4
- "eval_logits/rejected": -2.0897181034088135,
5
- "eval_logps/chosen": -33.78822326660156,
6
- "eval_logps/rejected": -37.54805374145508,
7
- "eval_loss": 0.49459660053253174,
8
- "eval_rewards/accuracies": 0.5921927094459534,
9
- "eval_rewards/chosen": 0.024633033201098442,
10
- "eval_rewards/margins": 0.027776462957262993,
11
- "eval_rewards/rejected": -0.0031434299889951944,
12
- "eval_runtime": 145.6868,
13
  "eval_samples": 343,
14
- "eval_samples_per_second": 2.354,
15
  "eval_steps_per_second": 0.295,
16
- "train_loss": 0.35194717815944127,
17
- "train_runtime": 10802.0809,
18
  "train_samples": 3079,
19
- "train_samples_per_second": 1.14,
20
- "train_steps_per_second": 0.143
21
  }
 
1
  {
2
+ "epoch": 1.0,
3
+ "eval_logits/chosen": -2.228182792663574,
4
+ "eval_logits/rejected": -2.2233598232269287,
5
+ "eval_logps/chosen": -34.08854293823242,
6
+ "eval_logps/rejected": -37.60850143432617,
7
+ "eval_loss": 24.80753517150879,
8
+ "eval_rewards/accuracies": 0.5132890343666077,
9
+ "eval_rewards/chosen": -0.005399257410317659,
10
+ "eval_rewards/margins": 0.0037888663355261087,
11
+ "eval_rewards/rejected": -0.009188123047351837,
12
+ "eval_runtime": 145.6001,
13
  "eval_samples": 343,
14
+ "eval_samples_per_second": 2.356,
15
  "eval_steps_per_second": 0.295,
16
+ "train_loss": 22.846377404943688,
17
+ "train_runtime": 3252.1936,
18
  "train_samples": 3079,
19
+ "train_samples_per_second": 0.947,
20
+ "train_steps_per_second": 0.118
21
  }
eval_results.json CHANGED
@@ -1,16 +1,16 @@
1
  {
2
- "epoch": 4.0,
3
- "eval_logits/chosen": -2.0943572521209717,
4
- "eval_logits/rejected": -2.0897181034088135,
5
- "eval_logps/chosen": -33.78822326660156,
6
- "eval_logps/rejected": -37.54805374145508,
7
- "eval_loss": 0.49459660053253174,
8
- "eval_rewards/accuracies": 0.5921927094459534,
9
- "eval_rewards/chosen": 0.024633033201098442,
10
- "eval_rewards/margins": 0.027776462957262993,
11
- "eval_rewards/rejected": -0.0031434299889951944,
12
- "eval_runtime": 145.6868,
13
  "eval_samples": 343,
14
- "eval_samples_per_second": 2.354,
15
  "eval_steps_per_second": 0.295
16
  }
 
1
  {
2
+ "epoch": 1.0,
3
+ "eval_logits/chosen": -2.228182792663574,
4
+ "eval_logits/rejected": -2.2233598232269287,
5
+ "eval_logps/chosen": -34.08854293823242,
6
+ "eval_logps/rejected": -37.60850143432617,
7
+ "eval_loss": 24.80753517150879,
8
+ "eval_rewards/accuracies": 0.5132890343666077,
9
+ "eval_rewards/chosen": -0.005399257410317659,
10
+ "eval_rewards/margins": 0.0037888663355261087,
11
+ "eval_rewards/rejected": -0.009188123047351837,
12
+ "eval_runtime": 145.6001,
13
  "eval_samples": 343,
14
+ "eval_samples_per_second": 2.356,
15
  "eval_steps_per_second": 0.295
16
  }
train_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
- "epoch": 4.0,
3
- "train_loss": 0.35194717815944127,
4
- "train_runtime": 10802.0809,
5
  "train_samples": 3079,
6
- "train_samples_per_second": 1.14,
7
- "train_steps_per_second": 0.143
8
  }
 
1
  {
2
+ "epoch": 1.0,
3
+ "train_loss": 22.846377404943688,
4
+ "train_runtime": 3252.1936,
5
  "train_samples": 3079,
6
+ "train_samples_per_second": 0.947,
7
+ "train_steps_per_second": 0.118
8
  }
trainer_state.json CHANGED
The diff for this file is too large to render. See raw diff
 
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b1aafdba9d2786e8fe506c065a1121076d97e18be040fc8361db9495c2d2dea5
3
  size 4984
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d1f31267c4e8a6a7c7100d15a541d9fbe9010773ed15e13f572c14b191f8b36
3
  size 4984